Remote plasma enhanced cyclic etching of a cyclosiloxane polymer thin film

The continuous evolution of chip manufacturing demands the development of materials with ultra-low dielectric constants. With advantageous dielectric and mechanical properties, initiated chemical vapor deposited (iCVD) poly(1,3,5-trimethyl-1,3,5-trivinyl cyclotrisiloxane) (pV3D3) emerges as a promising candidate. However, previous works have not explored etching for this cyclosiloxane polymer thin film, which is indispensable for potential applications to the back-end-of-line fabrication. Here, we developed an etching process utilizing O2/Ar remote plasma for cyclic removal of iCVD pV3D3 thin film at sub-nanometer scale. We employed in-situ quartz crystal microbalance to investigate the process parameters including the plasma power, plasma duration and O2 flow rate. X-ray photoelectron spectroscopy and cross-sectional microscopy reveal the formation of an oxidized skin layer during the etching process. This skin layer further substantiates an etching mechanism driven by surface oxidation and sputtering. Additionally, this oxidized skin layer leads to improved elastic modulus and hardness and acts as a barrier layer for protecting the bottom cyclosiloxane polymer from further oxidation.


Introduction
In contemporary chip fabrication, the integration of low-k dielectric materials in the interconnect layer plays a pivotal role in meeting the escalating demands of advanced chip technology [1,2].Materials possessing lower dielectric constants are essential for enhancing signal transmission speed and minimizing power consumption [3,4].Therefore, the 3 These authors contributed equally.* Author to whom any correspondence should be addressed.
Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.deployment of emerging low-k materials becomes imperative.The etching process serves as a key aspect for the integration of these materials particularly in achieving precise and controlled patterns.A tailored etching approach is essential to preserve the integrity of these materials and ensure the reliability and functionality of the intricate structures in modern semiconductor devices.
Among the emerging low-k materials, poly(1,3,5trimethyl-1,3,5-trivinyl cyclotrisiloxane) (pV 3 D 3 ) stands out as a promising candidate [4].PV 3 D 3 is a crosslinked organosilane polymer synthesized through free radical polymerization, where the presence of the cyclosiloxane ring structure contributes to the unique properties including a low-k constant.Extensive research has delved into the electrical [4], mechanical [5] and thermal conductive [6] properties of pV 3 D 3 .Previous studies have collectively highlighted its advantageous properties, showcasing pV 3 D 3 as a material capable of meeting the many stringent requirements including high dielectric strength and good stability [7,8].Despite these promising attributes, a comprehensive understanding of the etching process for pV 3 D 3 remains an underexplored area.While processes using O 2 plasma for surface modification of pV 3 D 3 have been reported [9], effective etching method has not been developed yet.
Dry etching of organosilicon films has been studied in the past, aiming for potential applications in bilayer lithography [10][11][12].Specifically, these siloxane polymer thin films can form SiO 2 surface layers that are resistant to O 2 reactive ion etching, thus serving as hard masks for pattern transfer.Most of these previous works were focused on polydimethyl siloxane [13,14] due to its simple molecular structure.As far as we know, highly crosslinked cyclosiloxane polymers have not been investigated because such structures cannot be easily formed by solution-based processes.In this work, we for the first time established a cyclic etching process aimed at layer-by-layer removal of initiated chemical vapor deposited (iCVD) pV 3 D 3 thin film.We employed quartz crystal microbalance (QCM) for monitoring the processes in situ and investigated the effect of process conditions on the etching rate.Surface analysis was also performed to study the possible mechanism of this process.

Deposition of pV 3 D 3 thin film
The pristine pV 3 D 3 thin film was deposited on Si wafer and QCM crystals in a custom-built iCVD reactor, which has been detailed elsewhere [6,15].The monomer of 1,3,5-trivinyl-1,3,5-trimethylcyclotrisiloxane (V 3 D 3 , Aladdin, 97%) and the initiator of tert-butyl peroxide (TBPO, J&K, 98%) were used as received without further purification.The V 3 D 3 monomer was heated at 70 • C to ensure an adequate vapor pressure and then delivered into the iCVD chamber at a flow rate of 0.9 sccm.The TBPO initiator, kept at room temperature, was delivered into the chamber at a flow rate of 0.6 sccm.The temperature of the iCVD chamber was maintained at 40 • C. The chamber pressure was controlled at 200 mTorr using a butterfly valve.The sample stage was kept at 30 • C by a circulating coolant, while the heated filaments positioned above the samples were set to 250 • C to cleave the initiators into free radicals.The growth rate of pV 3 D 3 film under this condition was ca. 30 nm•h −1 .

Remote plasma etching of pV 3 D 3 thin film
A customized plasma-enhanced atomic layer deposition (PE-ALD) system equipped with a QCM module was utilized for remote O 2 /Ar plasma etching (figure 1(a)).Specifically, the Cu coil mounted on a quartz tube serves as the inductively coupled plasma generator.The radio-frequency generator was operated at 13.56 MHz with an automatic matching network.The temperature of the chamber was set to 40 • C. Mass flow controllers were used to regulate the flow rates of the O 2 (99.999%) and the Ar (99.999%) gas.The dose of O 2 was controlled by an ALD valve.Ar was delivered separately as the carrier gas for plasma and the purging gas at fixed flow rates of 16 sccm, respectively.The chamber pressure was 43.2 Pa.
A typical etching cycle for pV 3 D 3 consisted of three steps.The first step included 10 s of O 2 dose followed by 10 s Ar purge.This step helped to release the pressure buildup of O 2 ahead of the ALD valve, therefore stabilizing the O 2 flow for the plasma.This O 2 dose step was also found important for obtaining a high etching rate.The second step ignited the remote plasma under a steady flow of O 2 and Ar, providing the oxygen radicals and Ar + ions for etching pV 3 D 3 .Finally, a long Ar purge (300 s) was performed to completely remove the volatile etching by-products after etching.

Characterization
We employed in-situ QCM for investigating the etching process.PV 3 D 3 thin film with thickness of ca. 30 nm was deposited onto the Au surface of the QCM crystals (6 MHz, Inficon 750-1057-G10).The coated crystal was then placed into the QCM housing in the center of the PE-ALD chamber (figure 1(b)).A Julabo circulator (CD-200F) was employed to further stabilize the temperature of QCM at 40.0 • C.
Ex-situ characterization was performed on the pV 3 D 3 coated Si coupons (ca. 2 × 2 cm) which were affixed around the QCM.The morphology of the pV 3 D 3 film was characterized by the atomic force microscope (AFM, Bruker Dimension XR) in tapping mode.The thickness of pV 3 D 3 film was measured by spectroscopic ellipsometry (J. A. Woollam, Alpha-SE).The as-obtained refractive index (n) and wavelength (λ) curve were subsequently parameterized by Cauchy dispersion model: where the Cauchy dispersion constants A = 1.493,B = 2.331 × 10 −3 and C = 2.054 × 10 −4 (mean squared error = 1.772) were determined using a pV 3 D 3 film with known thickness measured by AFM.Fourier-transform infrared (FTIR) spectra were measured in transmission mode, utilizing a Thermo Fisher Nicolet iS50 spectrometer equipped with a cryogenic HgCdTe (MCT) detector.X-ray photoelectron spectroscopy (XPS) was obtained from a Thermo Fisher ESCALAB Xi+ spectrometer.Nanoindentation experiments were conducted using a Nano Indenter G200 (Keysight Technologies).We employed focused ion beam (FIB, Helios G5 CX) to prepare the cross-sectional sample for transmission electron microscopy (TEM, Tecnai G2 F30) in both bright-field and annular darkfield scanning modes.Additionally, the element distribution was examined using energy dispersive x-ray spectroscopy (EDS).2(a).The mass loss after 30 cycles was found to be 1 191 ng•cm −2 , which corresponds to an average etching rate of 39.70 ng cm −2 •cycle −1 .Based on the loss of film thickness (20.20 nm) determined by ellipsometry, we found that the thickness-based etching rate is 6.73 Å cycle −1 .Notably, a positive mass change (78.17 ng•cm −2 ) was detected in the first cycle, which is likely ascribed to the surface oxidation of pV 3 D 3 due to exposure to oxygen-containing plasma [10].Without plasma ignition, O 2 doses alone were found In comparison with the process with O 2 /Ar plasma, the cyclic etching with Ar plasma led to over 10 times lower etching rate (0.62 Å•cycle −1 ) (figure 2(c)).This substantial decrease in etching rate indicates the importance of incorporating O 2 flow in plasma.The other noticeable difference between these two processes was the QCM response during and after plasma ignition.In the O 2 /Ar plasma process, we observed a quick mass gain followed with a sharp decline in mass during plasma ignition (figure 2(b)).During the subsequent Ar purge, a mass recovery was found in the first ca.30 s, followed with gradual mass loss till the end of the purge.This behavior could be attributed to the thermal effect on QCM crystal, parasitic voltages developed during the active phase of the plasma discharge and the re-deposition of etching by-products [16][17][18].In contrast, no such pronounced mass changes were detected during the etching cycles using solely Ar plasma (figure 2(d)).This result further substantiates that the combination of surface oxidation and ion sputtering is key to achieving high etching rates for pV 3 D 3 thin films.

Effect of process parameters on etching rate
We investigated the effect of plasma duration on etching rate.Figure 3(a) shows the mass change measured by in-situ QCM during the cyclic etching processes with plasma duration ranging from 5 s to 20 s.The plasma power was set to 300 W, while the O 2 flow rate was fixed at 10 sccm.We calculated the etching rate based on the mass loss in each cycle from the in-situ QCM data (figure 3(b)).We found positive mass changes (negative etching rates) during the first cycle with plasma duration of 5 s and 10 s, indicating that the surface oxidation extent in the first cycle in these process conditions was still not sufficient for etching.Even though we increased the plasma duration to 20 s to allow longer diffusion of reactive species into the polymer film, we only obtained mass loss of 1.96 ng•cm −2 during the first cycle, which further emphasizes the importance of the initial condition of surface oxidation state prior to the etching cycles.After the first cycle, the etching rates exhibit nonlinear behavior before reaching a steady state (marked with cross in the data points in figure 3(b)).These nonlinear behaviors stem from the competing mechanisms between surface oxidation and ion sputtering.Changes in the initial conditions of surface siloxane composition and the diffusivity of oxygen radicals into the film may also contribute to the nonlinear behaviors.Decrease in etching rates was also observed when the residue film thickness was close to zero.
We analyzed the etching rates during the cycles operating at steady state and found that these etching rates linearly scale with the plasma duration (figure 3(c)).The slope of the fitted line was found to be 5.52 ng These results indicate that the steady-state etching rate per unit time is independent of plasma duration.We used ellipsometry as a complementary characterization to determine the thickness change during the process.The thickness-based average etching rate was found to increase from 3.63 Å•cycle −1 -9.68 Å•cycle −1 with the rise of plasma duration from 5 s to 20 s.
We further studied the effect of plasma power on etching rate.As shown in figure 4(a), the mass loss after 30 cycles of etching increased as the plasma power was raised from 50 W to 300 W and slowly saturated with further increase of plasma power.Based on the etching rates calculated from the in-situ QCM data (figure 4(b)), we found that only conditions with plasma power ranging from 50 W to 300 W reached steady state during the first 30 cycles.The averaged steady state etching rates (ER ss , in units of ng•cm −2 •cycle −1 ) exhibit a linear dependence on the plasma power (figure 4(c)).We performed a linear fitting to the following empirical equation where the slope φ was found to be 0.217 ng•cm −2 •cycle −1 • W −1 , P is the plasma power (W), P th is the power threshold found to be 46.0W.
We also studied the effect of O 2 flow rate on etching rate (figure 5(a)).As discussed above, the absence of O 2 flow led to very low etching rate (figure 5(b)).Increasing the O 2 flow rate from 0 to 5 sccm resulted in dramatic improvement of the steady state etching rate from 3.47 ng•cm −2 •cycle −1 -55.1 ng•cm −2 •cycle −1 (figure 5(c)).The etching rate saturated at ca. 55 ng•cm −2 •cycle −1 even though the O 2 flow rate was further enlarged.However, we noticed that the steady state appeared in earlier cycles in the conditions with 10 sccm and 20 sccm of O 2 flow than the process with 5 sccm of O 2 .Consequently, larger mass loss was achieved after 30 cycles with 10 sccm and 20 sccm of O 2 flow.Ellipsometry results show that the average etching rates for the processes with 10 sccm and 20 sccm of O 2 flow are 6.73 Å•cycle −1 and 7.28 Å•cycle −1 , respectively.These etching rates are substantially higher than the process with 5 sccm of O 2 (5.44 Å•cycle −1 ), which agrees well with the in-situ QCM data.before and after etching was 170.54 nm and 151.10 nm, respectively.For the pristine pV 3 D 3 thin film, the peaks at 740 cm −1 and 1003 cm −1 are associated with symmetric and asymmetric stretching of Si-O-Si, respectively [19,20].The vibration modes at 795 cm −1 for Si-CH 3 rocking [21,22], 970 cm −1 for CH 2 wagging in pendant vinyl groups and 1260 cm −1 for the Si-CH 3 symmetric bending [3,23] all match well with the literature.The spectrum of the etched pV 3 D 3 film resembles that of the pristine film, indicating that the bulk structure of the pV 3 D 3 film was not altered during etching.

Structural characterization and etching mechanism
We further utilized XPS to investigate changes on the surface attributable to the etching process.The substantially enhanced surface oxygen content (figure 6(b)) provides direct evidence for the surface oxidation of pV 3 D 3 .We deconvolved the C 1 s spectra (figure 6 We also found the conversion of siloxane structures from di-functional (D) units to quad-functional (Q) units during the etching process.The Si 2p spectrum of the pristine pV 3 D 3 shows a single peak at 102.5 eV, corresponding to the Si-O bonds within the D3 ring [27].This peak became asymmetric and shifted to higher binding energy after etching due to the increase of oxygen content in the SiO x C y structure [28,29].The deconvolution for the Si 2p spectrum of the etched pV 3 D 3 revealed a new peak at ca. 103.6 eV (figure 6(d)), confirming the formation of SiO 2 -like Q units on the surface according to previous studies [29][30][31].This surface layer was further confirmed by FIB and cross-sectional TEM.The thickness of this layer was found to be ca.7.5 nm (figure 6(e)).Moreover, figure 6(g) reveals that the Si concentration in the surface layer is higher than that in the pV 3 D 3 film, which is consistent with our XPS analysis.
Based on the above discussion, we proposed a mechanism for our etching process.As illustrated in figure 7(a), the O• radicals oxidize the surface of pV 3 D 3 , leading to fragmented chains and the transformation of cyclosiloxanes into SiO 2 -like Q units (figure 7(b)).This oxidized skin layer is similar to what has been reported for O 2 plasma etching of polydimethyl siloxane films [10].The oxidized skin layer is likely removed by Ar + ions and excited Ar atoms.Notice that no substrate bias was applied in our cyclic etching process.However, Ar + and excited Ar atoms may still possess a certain kinetic energy upon reaching the substrate surface [32,33].The O• radicals replenished in each etching cycle keep oxidizing the pV 3 D 3 surface, enabling continuous cyclic etching.As the skin layer recedes during the etching process, this layer also acts as a barrier layer for protecting the bottom organosiloxane film from further oxidation [12].
Finally, we investigated the changes of morphology and mechanical properties for pV 3 D 3 thin film before and after etching.The AFM images in figures 8(a) and (b) compare the morphology of the pristine and etched pV 3 D 3 thin film.The surface roughness of the pristine film was 0.587 nm, which slightly reduced to 0.474 nm after the etching.This improved smoothness would be important for controlling line edge roughness during patterning transfer.
Figure 8(c) shows the load-displacement curves of pristine and etched pV 3 D 3 thin films obtained by continuous stiffness measurement method.The hardness of the thin films was derived as a function of indent depth.Averaged from the displacement range from 15 nm to 20 nm (ca.10% of the film thickness), the hardness of the pristine film ((0.39 ± 0.01) GPa) was found similar to previously reported values (ca.0.4 GPa) [6,15].The etched pV 3 D 3 thin film exhibits a higher hardness ((0.45 ± 0.01) GPa) (figure 8(d)).This slight increase in film hardness can be explained by the SiO 2 -like units in the skin layer formed during the etching process, as SiO 2 possesses significantly higher hardness (>10 GPa) [34] than pV 3 D 3 .In addition, we found that the Young's modulus was also increased from (6.38 ± 0.11) GPa to (6.94 ± 0.14) GPa after 30 cycles of etching.

Conclusions
We have developed a novel cyclic etching method using O 2 /Ar remote plasma and achieved layer-by-layer removal of pV 3 D 3 thin film.Our findings indicate that the etching rate can be modulated by adjusting the plasma duration, plasma power and O 2 flow rate.The surface oxidation of pV 3 D 3 by O• radicals is critical for obtaining a high etching rate.The formation of an oxidized skin layer was confirmed by in-situ QCM and ex-situ XPS results, which acts as a barrier for protecting the underlying pV 3 D 3 thin film during etching.The surface roughness decreases after etching, accompanied with a slight increase in hardness and Young's modulus, attributable to the oxidization of the film's surface.

Figure 1 .
Figure 1.The custom-designed PE-ALD system equipped with a QCM module.(a) Schematic diagram of the PE-ALD system used for remote plasma etching of pV 3 D 3 .The unused ALD components are omitted for clarity.(b) Top view of the PE-ALD chamber equipped with an in-situ QCM module.

Figure 2 .
Figure 2. In-situ QCM data for cyclic etching of pV 3 D 3 thin film with O 2 /Ar plasma and with Ar plasma alone.(a) In-situ QCM data for cyclic etching of pV 3 D 3 thin film with O 2 /Ar plasma.Plasma power was set to 300 W, while plasma duration in each cycle was 10 s.(b) A representative cycle during the O 2 /Ar plasma etching process.The O 2 dose step is marked in blue, while the O 2 /Ar plasma step is marked in pink.(c) In-situ QCM data for cyclic etching of pV 3 D 3 thin film with Ar plasma (no O 2 flow).(d) A representative cycle during the Ar plasma etching process.The Ar plasma duration is marked in pink.

Figure 2 (
Figure 2(a) shows the mass change measured by in-situ QCM during the cyclic etching process with 10 s of O 2 /Ar remote plasma in each cycle.The plasma generator operated at a power setting of 300 W, with gas flow rates set to 10 sccm for O 2 and 16 sccm for Ar.The successful layer-by-layer removal of pV 3 D 3 thin film was clearly observed in figure2(a).The mass loss after 30 cycles was found to be 1 191 ng•cm −2 , which corresponds to an average etching rate of 39.70 ng cm −2 •cycle −1 .Based on the loss of film thickness (20.20 nm) determined by ellipsometry, we found that the thickness-based etching rate is 6.73 Å cycle −1 .Notably, a positive mass change (78.17 ng•cm −2 ) was detected in the first cycle, which is likely ascribed to the surface oxidation of pV 3 D 3 due to exposure to oxygen-containing plasma[10].Without plasma ignition, O 2 doses alone were found

Figure 3 .
Figure 3.Effect of plasma duration on etching rate.(a) Mass change measured by in-situ QCM during the cyclic etching process with varied plasma duration.The plasma power was set to 300 W, while the O 2 flow rate was 10 sccm.(b) Etching rate calculated from in-situ QCM data.Data points marked with cross ('+') represent the cycles operating at steady state.(c) Etching rate averaged from the cycles operating at steady state.The dashed line represents a linear fitting.

Figure 4 .
Figure 4. Effect of plasma power on etching rate.(a) Mass change measured by in-situ QCM during the cyclic etching process with varied plasma power.The plasma duration was set to 10 s, while the O 2 flow rate was 10 sccm.(b) Etching rate calculated from in-situ QCM data.Data points marked with cross ('+') represent the cycles operating at steady state.(c) Etching rate averaged from the cycles operating at steady state.The dashed line represents a linear fitting.

Figure 6 (
Figure 6(a) shows the transmission FTIR spectra of both pristine and etched pV 3 D 3 thin films.The film thickness

Figure 5 .
Figure 5.Effect of O 2 flow rate on etching rate.(a) Mass change measured by in-situ QCM during the cyclic etching process with varied O 2 flow.The plasma power was fixed at 300 W, while the plasma duration was set to 10 s.(b) Etching rate calculated from in-situ QCM data.Data points marked with cross ('+') represent the cycles operating at steady state.(c) Etching rate averaged from the cycles operating at steady state.
(c)) according to previous references to investigate the change of C contents on the surface [24-26].The rise of C-O and C = O (ca. 286.6 eV and 289.0 eV, respectively), along with the decay of C-C (ca.284.8 eV), is likely attributed to the oxidation of the methyl groups and pendant vinyls on the D3 ring as well as the polymer backbones of pV 3 D 3 .

Figure 6 .
Figure 6.Changes in the surface composition of pV 3 D 3 thin film after etching.(a) FTIR spectra of pV 3 D 3 thin film before and after etching.(b) Surface composition obtained from XPS spectra.(c) C 1 s and (d) Si 2p XPS spectra for pV 3 D 3 thin film before and after etching.The cross-sectional (e) bright-field and (f) dark-field TEM images and (g) EDS mapping image for Si distribution in the pV 3 D 3 thin film after etching.

Figure 7 .
Figure 7. Schematic illustration of the proposed etching mechanism.(a) Proposed etching mechanism.The oxidized skin layer is formed via the reaction between pV 3 D 3 and O• radicals, which is likely removed by Ar + ions and excited Ar atoms.(b) Proposed surface oxidation mechanism.Di-functional siloxane units are converted to quad-functional units that are close to SiO 2 .

Figure 8 .
Figure 8. Morphology and mechanical properties of pV 3 D 3 thin films before and after etching.(a)-(d) AFM images of the pristine pV 3 D 3 thin film (a) before and (b) after 30 cycles of etching.(c) Mechanical response measured by nanoindentation.(d) Hardness comparison between pristine and etched pV 3 D 3 films.